Method for Providing a Perfusion Based Complication Prediction and Clinical Decision Support System

The present application claims priority to European Patent Application No. EP 25 165 062.8, filed on Mar. 20, 2025, wherein the entire contents of which are incorporated herein by reference for all purposes.

The present disclosure relates to a method and a system for providing a perfusion based complication prediction, a clinical decision support system, and a computer program.

Colonic ischemia (CI) is a complication that can occur after abdominal aortic aneurysm (AAA) repair surgery. This complication significantly affects the morbidity and mortality of this type of surgery. Often, this complication will only be detected days or weeks after the surgery has been performed. Earlier detection of CI can help reduce the morbidity and mortality of AAA repair surgery.

ICG (indocyanine green) fluorescence imaging is frequently used in evaluating perfusion of organic tissues. An example is colonic perfusion examined during abdominal aortic aneurysm (AAA) repair. During AAA repair, the aorta is clamped to prohibit blood flow through the aorta while the damaged aorta is repaired. After completion of the repair of the aorta, the clamp is removed and blood is once again allowed to flow freely through the aorta.

In the past, it has been known for the surgeon to evaluate colonic perfusion using ICG fluorescence by visual inspection or using non-standardized reference measurements, which does not yield reliable outcomes. One reason for this is that state-of-the-art methods for objective assessment of ICG fluorescence images often use generic cut-off values irrespective of the individual patients, making objective assessment unreliable, because fluorescence data can vary greatly between patients. This depends on multiple factors, including but not limited to camera settings, distance of the camera, such as an exoscope, to the surgical site and also on the patient himself. The latter in turn depends on how well the patient is able to adapt to impaired perfusion, e.g. by collateral perfusion.

In light of this, it is an object of the present disclosure to provide means for objective assessment of the risk of complications due to perfusion issues in a patient.

1 2 3 1 2 3 a) ICG perfusion image time series of a patient's organ receiving blood supply from the operated on blood vessel from times T, Tand Tare loaded from a database, Tdenoting a time before clamping shut the blood vessel, Tdenoting a time after fully clamping shut the blood vessel and before repair of the blood vessel and Tdenoting a time after finishing the repair of the blood vessel and removing the clamp, 1 2 3 b) the T, Tand Tperfusion image time series are normalized to a common standard and ICG fluorescence intensity time series extracted for a plurality of Regions of Interest on an organ supplied by the vessel, 1 2 c) a set of one or more collective characteristics of the fluorescence intensity time series at Tand Tare taken as benchmarks for unobstructed and maximally obstructed blood supply, respectively, 3 1 2 d) the same set of collective characteristics of the fluorescence intensity time series at Tis evaluated with respect to its collective proximity to those of the perfusion image time series at Tand T, respectively, the proximity to either being taken as an indicator for the complication prediction, and e) a complication prediction based on the outcome of method step d) is output. This object can be achieved in an aspect of the present disclosure by a method for providing a perfusion based complication prediction for blood vessel related operation outcomes, the method being carried out post-operatively or during an operation, wherein

The method may not include surgical steps to be performed on the patient during an operation, the surgery itself and the administration of ICG into the patient.

This evaluation forms the basis of an objective patient based perfusion assessment model for assessing colonic or other organic perfusion during or after surgery such as open abdominal aortic aneurysm (AAA) repair surgery. The model can aid the surgeon in making an informed decision about the patient's risk of developing complications such as colonic ischemia or in other cases where perfusion assessment can be critical for patient outcomes.

The ICG fluorescence time series extracted for the plurality of Regions of Interest constitute normalized time intensity curves extracted from the ROIs or from the pixel or a combination of pixels from the ROIs.

To make the evaluation more robust, it can be envisioned to additionally provide a region of interest on a different organic structure having less perfusion, such as the mesentery. Additionally, or alternatively, it can be also possible to use a piece of cloth to establish a baseline in the ICG fluorescence time series.

The ICG perfusion image time series may be open surgery near infrared (NIR) videos taken while ICG fluorescent marker can be injected into the patient and show the spread and subsiding of ICG fluorescence in the perfused organs such as the colon. The time series may last for half a minute, a minute or two minutes.

average, maximum and/or variance of the ingress slope; average, maximum and/or variance in intensity during the ingress phase; average, maximum and/or variance of area under the curve (AUC) of ingress phase; average, maximum and/or variance of slope in the first 10 seconds after start of ingress; average, maximum and/or variance in intensity in the first 10 seconds after start of ingress; average, maximum and/or variance of area under the curve (AUC) of first 10 seconds after start of ingress; and average, maximum and/or variance of egress slope. After processing the time intensity curves, time intensity curve parameters can be calculated based on these curves. The number and selection of parameters can vary from patient to patient or from procedure to procedure. The selected parameters can include, but do not have to be limited to:

1 2 3 The terms average, median, maximum and variance refer to the collective of these parameters, since the curves of all regions of interest can be evaluated collectively and their average, maximum or median values can be used. The reason for the collective analysis of multiple ROIs can be that no single ROI can be reliably identified and relied upon to yield dependable results and having dependable predictive value. On the other hand, since the analysis can be a collective one, it can be possible to define different regions of interest in the ICG image time series at T, Tand T, as long as they can be similarly distributed, which makes movements from one time series to the other less of an issue with respect to comparability.

1 2 3 1 2 3 3 3 3 1 3 2 1 2 1 2 3 2 In a further embodiment, the collective characteristics from the T, Tand Ttime series can be scaled such that the value from Tcan be positive, such as for example +1, the value from Tcan be negative, such as for example −1, and the value from Tcan be scaled accordingly. If multiple collective characteristics can be used, a mean ratio of the Tvalues may be calculated from the scaled Tcollective characteristics. A positive mean ratio will show a better resemblance of the Tdata to the Tdata and therefore stand for the preponderance of likelihood of a positive, complication free patient outcome, whereas a negative mean ratio will show a better resemblance of the Tdata to the Tdata and stand for the preponderance of likelihood of a negative patient outcome having complications. In other words, as the Tvideo was taken before the AAA repair and the Tvideo was taken after clamping the Aorta, the Tvideo shows proper aortic blood supply to the colon and the Tvideo shows impaired aortic blood supply to the colon. Therefore, if the Tvideo shows a closer resemblance to the Tvideo, this can be an indication of impaired aortic blood supply to the colon.

After placing the regions of interest, the ICG fluorescence intensity time series may be extracted by calculating the average NIR intensity for each region of interest for each frame. Once the ICG fluorescence intensity time series can be extracted, the curves may be processed in further embodiments by subjecting the ICG perfusion image time series to image processing, wherein for each ICG perfusion image time series, the signal curves can be filtered, normalized, and the baseline value can be subtracted. The videos, i.e., the ICG perfusion image time series, may also be registered to remove most of the movement within the time series, or motion-tracking may be used on the ROIs to compensate for movements. This helps increase the quality of the extracted signal curves, i.e., the extracted ICG fluorescence intensity time series.

1 2 1 2 1 2 1 2 In further embodiments, the collective characteristics chosen for evaluation can be determined beforehand based on Tand TICG fluorescence intensity time series extracted from ICG perfusion image time series of a plurality of patients and performing a statistical test on the extracted Tand TICG fluorescence intensity time series designed to identify those collective characteristics that show the most significant difference between their Tand Tvalues. The collective characteristics having the most distinctive difference between their Tand Tvalues can be most indicative of impaired blood supply under clamping. This comparison may be done, e.g., using the Wilcoxon signed-rank test, a paired statistical test, or the t-test for matched pairs if the sample size can be large enough.

In embodiments, at least one ingress-related parameter and at least one egress-related parameter may be selected for use in a decision model. Besides or in place of these parameters, the variance between the parameter values for each of the videos may also calculated. The variance of a parameter, such as an ingress-related parameter, may be selected as another parameter for the decision model.

The object underlying the present disclosure can be also achieved by a computer-based Clinical Decision Support System (CDSS) capable of providing a perfusion based complication prediction for blood vessel related operation outcomes, the system comprising a computer connected to or comprising a database and running a software program configured to perform the above-described method.

With this, the CDSS embodies similar ideas, characteristics, features and advantages as the method.

1 2 3 In embodiments, the CDSS further comprises a camera sensitive to the ICG fluorescence spectrum, an illumination device configured to provide excitation illumination for ICG, and a controller controlling the camera and the illumination device, wherein the camera can be configured to capture ICG fluorescence image time series and the controller can be configured to store the ICG fluorescence image time series in the database along with identifying information and information about the captured ICG fluorescence image time series being a T, Tor Ttime series.

1 2 3 1 2 3 1 2 3 In further embodiments, a method of operating on a patient may include capturing T, Tand TICG perfusion image time series on a patient, Tdenoting a time before clamping shut the blood vessel, Tdenoting a time after fully clamping shut the blood vessel and before repair of the blood vessel and Tdenoting a time after finishing the repair of the blood vessel and removing the clamp, and analysing the T, Tand TICG perfusion image time series according to the above-described method in order to arrive at a perfusion based prediction for blood vessel related operation outcomes.

10 In a further aspect, the object of the present disclosure can be also achieved by a computer program with software code that cause a computer of a computer-based CDSS (), as described above, to perform the steps of the above-described method.

1 2 3 1 2 3 1 2 3 Also envisioned can be a method for surgery on a patient that involves clamping shut and repairing a blood vessel, such as an abdominal aortic aneurysm (AAA) repair surgery, comprising administering ICG to the patient at times T, Tand Tand capturing ICG perfusion image time series of a patient's organ receiving blood supply from the operated on blood vessel at the times T, Tand T, with Tdenoting a time before clamping shut the blood vessel, Tdenoting a time after fully clamping shut the blood vessel and before repair of the blood vessel and Tdenoting a time after finishing the repair of the blood vessel and removing the clamp, and carrying out a perfusion based complication prediction for blood vessel related operation outcomes according to the above-described method of the present disclosure.

1 2 3 The T, Tand TICG perfusion image time series may be stored in a database for analysis.

Further characteristics of the disclosure will become apparent from the description of the embodiments according to the disclosure together with the claims and the included drawings. Embodiments according to the disclosure can fulfil individual characteristics or a combination of several characteristics.

In the drawings, the same or similar types of elements or respectively corresponding parts are provided with the same reference numbers in order to prevent the item from needing to be reintroduced.

1 FIG. 10 10 12 12 18 14 16 14 18 14 16 illustrates a schematic representation of a computer-based Clinical Decision Support System (CDSS)that can be equipped and configured to carry out the method of providing a perfusion based complication prediction for blood vessel related operation outcomes. The CDSShas a central computerrunning the software for the method. The computercan be connected to a controllerthat controls a camera, such as an exoscope, and an illumination devicethat can be configured to provide illumination light for different illumination modes such as white light and infrared lighting for fluorescence excitation of the ICG fluorescence agent. The cameracan be configured to capture image sequences, that is, videos, in the visible spectrum and in a fluorescence capture mode, or in the near infrared spectrum. The controllercan be configured to control the capturing mode of the cameraand the illumination mode of the illumination deviceto conform to white light imaging or fluorescence imaging, respectively, as the case may be.

12 18 18 14 16 12 14 20 12 22 The central computercan be in communication with the controllerand may be configured to instruct the controllerto perform control of the cameraof the illumination deviceaccording to the image capturing mode required by the circumstances of a procedure. The computerreceives the captured image sequences from the cameraand stores them in a Database. Furthermore, the computercan be connected to a user interface device, such as a monitor, keyboard, mouse or the like.

2 FIG. 100 14 16 10 1 2 3 can be a schematic representation of a methodfor providing a perfusion based complication prediction for blood vessel related operation outcomes. It can be assumed that, for example during an abdominal aortic aneurysm (AAA) repair surgery on a patient, fluorescence videos, i.e., ICG perfusion image time series have been taken by means of cameraand illumination deviceof the CDSS, where an ICG fluorescence agent can be injected during the operation at various times. During the operation, such ICG perfusion imaging time series can be taken at a time Tbefore clamping shut the aorta, then at a second time Tafter clamping shut the aorta, and then at a third time T, which can be after the AAA repair surgery has been completed and the clamping of the aorta has been removed. In each case, the videos show an organ whose blood supply comes mainly from the aorta, such as the colon. Other organs may, as the case may be, the ileum, the gastric conduit, trachea, the parathyroid gland among others.

In the case of AAA, the clamping shut of the aorta will have an effect on the perfusion of the colon with blood and hence with the ICG fluorescence agent. This agent can be injected intravenously, usually the arm. However, in the aortal repair, patients often have central IV lines through the neck veins to the larger veins near the heart. Especially in the latter case, distribution of the ICG fluorescence agent can happen very quickly.

102 12 1 2 3 20 104 1 2 3 In stepof the method, the computerloads the T, T, TICG perfusion image time series from the Databasefor further processing. A first processing stepmay comprise normalizing the T, T, TICG perfusion image time series. This step may also comprise filtering and subtracting of baseline values.

106 1 2 3 1 2 3 In step, multiple regions of interest can be defined on the observed organ in the respective ICG perfusion imaging time series. Further regions of interest may be defined on control regions such as the mesentery, or a piece of cloth that does not have the ICG in it and therefore can serve as a baseline observation to react to changes in lighting. An equal number of regions of interest can be defined in each of the T, Tand TICG perfusion imaging time series, but this can be not a strict requirement. The definition of multiple regions of interest helps in avoiding the results being defined by a single or just a few, possibly nonrepresentative, example or examples. Also, the regions of interest need not be the same in all three ICG perfusion imaging time series, since from time Tto time Tto time Tthe patient or the camera may have been moved so that the organ under observation can be located differently in all three videos and it can be impossible or impractical to try to define the identical regions of interest over all three times. This impracticability can be alleviated by the choice of multiple regions of interest in each of the three ICG perfusion imaging time series that provide statistical robustness to the analysis.

108 1 2 3 In step, T, Tand TICG fluorescence intensity time series can be extracted from the regions of interest that have been defined earlier. The normalization and baseline subtraction mentioned earlier may also or alternatively be applied to the ICG fluorescence intensity time series at this stage instead of earlier with respect to the ICG perfusion imaging time series. The difference can be that, whereas the ICG perfusion imaging time series represent image series, that is, videos, the ICG fluorescence intensity time series represent only one value per region of interest per timeframe, namely the intensity of the fluorescence response.

1 2 3 The ICG fluorescence intensity time series of a single region of interest will therefore be a time dependent graph or curve that in many cases shows a steep rise during ingress of the ICG and a relatively flat decrease during the time that the ICG can be metabolized through the liver, the so-called egress phase, whereas the ICG fluorescence intensity time series of all regions of interest combined will be a bundle of such time dependent curves that will be spread out somewhat, enabling a statistical analysis thereof including analysis of minimum, maximum, average values or their characteristic distributions such as variance, full width half maximum (FWHM) or the like. Normalization may mean that the maximum intensity value in any of the regions of interest can be defined as, for example, 100%, with the scaling factor to this 100% value being taken as a normalizing factor and all the curves for all the regions of interest can be normalized with that normalizing factor. For each of the T, Tand TICG fluorescence intensity time series, the normalization may be done individually. However, if it can be ensured that lighting and positioning for the videos at the three different times did not change, the normalization may be performed with the same normalization factor for all three ICG fluorescence intensity time series. Furthermore, a baseline subtraction may also be performed with respect to the curves on the basis of the region of interest defined on a neutral part of the images, such as a piece of cloth or as a gray card placed in a non-obstructing part of the images.

110 1 2 3 In the next step, collective characteristics can be extracted from the T, T, TICG fluorescence intensity region of interest time series. This means, that the bundle of intensity curves can be analysed with respect to certain collective characteristics of the curve overall. Such characteristics may be determined for different phases of the curves, such as during ingress, up to maximum or a pre-defined time after start of ingress, such as 10 seconds, 20 seconds and so on, or during egress. Such characteristics may, e.g., be the slope, the value or the area under the curve during the various different phases. The collectiveness aspect may be realized by determining a maximum, average, minimum, variance or such of the characteristic under consideration for all the curves of the bundle from the multiple regions of interest. The analysis will usually be done for a number of such collective characteristics that have been shown to be most indicative of positive or negative patient outcomes.

112 3 1 2 1 2 3 1 2 1 114 In step, a proximity or proximities of extracted Tcollective characteristics to the respective extracted Tand Tcollective characteristics can be evaluated, where the Tand Tcollective characteristic values represent the individual, i.e. subjective to the patient, ideal and worst-case values. The Tcollective characteristic values will ideally resemble the Tvalues for a positive outcome. If they resembled the Tvalues more than the Tvalues, this may indicate a possibly negative patient outcome. This can be calculated in stepaccording to a weighting model and a complication prediction can be output to the user.

3 FIG. 200 100 1 2 202 1 2 20 102 104 106 108 110 100 1 2 1 2 shows a schematic representation of a methodfor choosing collective characteristics that can be used in method. The task at hand can be basically defined those characteristics that, for a larger number of patients, provided the starkest distinction between the Tand Tstates. For this purpose, in a first step, pre-normalized Tand TICG fluorescence intensity time series for multiple patients constituting a data pool can be loaded from database. Alternatively, one or more of the steps,,,andof methodcan be carried out for Tand TICG perfusion imaging time series from a sufficiently large number of patients undergoing the same or similar procedures, thus creating the plurality of normalized of Tand TICG fluorescence intensity time series for the patients from the data pool.

1 2 204 206 1 2 2 1 2 1 2 Next, a variety of collective characteristics can be extracted from the Tand TICG fluorescence intensity time series in stepin a uniform fashion for all patients in the data pool, thus ensuring comparability and the ability to perform a statistical analysis in stepthat can be designed to determine those collective characteristics that can be most significantly different for Tand Tfluorescence intensity time series, respectively. For example, in case of AAA repair surgery, the TICG fluorescence intensity time series can be compared to the TICG fluorescence intensity time series for each patient in order to assess which parameters can be indicative of impaired aortic blood supply to the sigmoid colon, as the Tvideos all have impaired aortic blood supply to the sigmoid colon due to the clamping. This comparison may for example be done using the Wilcoxon signed-rank test, a paired statistical test. Since there may not be a direct relation between the plurality of regions of interest from the Tand the TICG perfusion imaging time series from the same patient, a single value for ICG perfusion imaging time series, i.e., video, will have to be calculated in order to perform the statistical test. To do this, the median value of each parameter from the plurality of regions of interest may be calculated for each video, resulting in a single value per parameter per video.

208 1 2 In step, the resulting collective characteristics definitions can be output that provide the most reliable discrimination power between the Tand the Tcharacteristics.

1 2 In an exemplary embodiment, the statistical test results may be used to assess which parameters show a significant difference in values between Tand T. From those parameters, for example, one ingress-related parameter and one egress-related parameter may be selected for use in a decision model. Besides these parameters, the variance between the parameter values for each of the videos may also be calculated. The variance of an ingress-related parameter may for example be selected as the third parameter for the decision model. Other choices may be made as well.

100 1 2 3 Although the determination of the most distinguishing collective characteristics can be done on a plurality of patient's ICG fluorescence intensity time series, the evaluation in methodof the individual risk of complications can be done purely subjectively, that is, using only the individual patient's T, Tand TICG fluorescence intensity time series data.

2 FIG. 112 100 3 2 1 3 2 1 3 2 1 2 1 3 3 1 2 3 2 Referring again to, in stepof method, for the evaluation of the likelihood of complications of an individual patient, the value of that patient's Tmeasurement would be compared to the values of the Tand Tmeasurements for that individual patient for each of the chosen parameter. It can be then assessed whether the Tmeasurement value can be closer to the Tmeasurement value or to the Tmeasurement value, and by how much. The Tvalue can be scaled based on the Tand Tvalues, to make the model quantitative and allow for better comparison between patients and parameters. Here, the Tvalue may be defined as −1, and the Tvalue may be defined as +1. The Tvalue can be then scaled accordingly. Afterwards, the mean value of the three Tvalues (one for each parameter) can for example be taken to get a final value. Other weighting factors may be used based on the differing statistical significance of the various collective characteristics. Given the definition of the Tand Tvalues as +1 and −1, if this value can be negative, the Tmeasurement more closely resembled the Tmeasurement, indicating impaired aortic blood supply to the sigmoid colon. If this value can be positive, the opposite can be true.

4 FIGS. 4 FIG. 1 2 3 1 2 3 1 2 3 30 1 32 2 34 3 In, T, Tand TICG fluorescence intensity time series derived from videos of the sigmoid colon during AAA repair surgery of a first patient and a second patient can be shown as representative examples, with the data of the first patient in the upper half and the data of the second patient in the lower half of. In both cases, the T, Tand TICG fluorescence intensity time series for all of the chosen regions of interest can be displayed, with the horizontal axis representing a time axis spanning approximately two minutes from the start of the injection of the ICG fluorescence agent into the patient. The T, Tand TICG fluorescence intensity time series can be shown as bundles of curves that can be individualized with reference numberin the case of the Tseries, reference numberin the case of the Tseries and reference numberin the case of the Tseries. The bundles of curves have been normalized such that, for each bundle, the maximum ICG fluorescence intensity can be scaled to the value of 100. This also means, that only one of the curves of each bundle will reach the maximum value of 100.

36 1 2 3 3 1 2 Furthermore, a negative control can be highlighted with reference number. Also, in each case there has been one region of interest been defined on the mesentery. The T, Tand Tmesentery curves can be shown in a dashed line. These usually lie within the group of curves for the regions of interest on the sigmoid colon. They have not been used in the statistical analysis and the comparison of the Tvalues with the Tand Tvalues.

4 FIG. The ICG fluorescence intensity time series shown as curves inhave been subjected to some signal processing. These include filtering, wherein the curves were filtered using a Whittaker-Henderson filter with a lambda value of 500 k. This filtering helped remove unwanted noise in the signal. Furthermore, a baseline subtraction was applied, wherein a baseline value was subtracted from each signal. The baseline value can be defined as the average intensity of the curve before the average NIR value in the whole frame reaches >1% of its maximum intensity. Finally, the signals were normalized using a combined normalization method, wherein all signals of a video can be scaled together so that the maximum value of the highest curve was at 100.

4 FIG. 1 2 2 1 3 1 2 1 As can be seen in the upper half of, the Tcurves of the first patient start to rise early after the injection, reach a well-defined maximum and starts to decrease to a lower level very quickly, where the lower level can be around 50, that is, half of the maximum, and then decrease very slightly. The Tcurves show the impact of the clamping of the aorta. The rise during ingress of the ICG fluorescence agent starts later, can be less steep and does not show a pronounced peak. Furthermore, the spread, that is, the variance, of the curves of the Tcurves, can be much larger than in the Tcase. The Tcurves resemble the Tcurves much more than the Tcurves in that they also rise quickly and present a pronounced maximum before going back to a reduced value of about 40% of the maximum. The spread can be somewhat larger than in the Tcase. In this case, the decision model has predicted that the aortic blood supply to the sigmoid colon will not be impaired. The patient outcome was indeed such that no ischemia was observed.

4 FIG. 1 2 1 1 2 32 3 1 3 2 In the lower half of, it can be seen that the Tand Tcurves of the second patient differ from those of the first patient. The Tcurve indeed shows a rather quick rise during ingress and a maximum, wherein however the falloff can be a little less pronounced than in the first patient. Also, the curves in the Tbundle show a greater spread than in the case of the first patient. The Tcurves show a very different behaviour. Most of the curves that can be indicated by the ellipse with the reference numberrise very slowly, and they show a very large spread. Within the two minutes observation time, no peak can be observed and all of the curves can be still rising. In the case of the Tcurves, there can be once again a quick rise, but there can be no pronounced maximum and the curves do not fall off significantly after the maximum can be reached. In this case, although the initial rise has a steepness that can be comparable to that of the Tcurves, the Tcollective characteristic values can be actually closer to the Tcollective characteristic values so that the prediction can be that the aortic blood supply to the sigmoid colon will be impaired. In this case, the patient suffered ischemia, proving the prediction correct.

5 FIG. 1 2 3 1 2 1 2 1 2 3 1 2 3 2 1 illustrates outcomes of a selected characteristic at T, Tand Tfor a plurality of patients. The selected collective characteristic can be the maximum slope during the ingress phase, that can be the slope at its steepest point on the rising portion of the curves. For each patients in the data pool, the Tand Tvalues for this collective characteristic can be displayed as endpoints of a line connecting the Tand Tvalues. In all cases, the Tvalue can be larger than the Tvalue. In this case, no normalization to the range from −1 to +1 has been applied, the slope values can be represented as percent per second (%/s). The respective Tvalue can be represented as a point that either lies on the line between the Tand Tvalues or outside the line, if the Tvalue can be smaller than the Tvalue or larger than the Tvalue.

8 9 1 2 8 9 1 2 4 FIG. In the case of patientsand, which can be the same as those whose ICG fluorescence intensity time series curves can be shown in, the collective characteristic lie above and, respectively, below the Tand Tvalues, giving patienta good outlook and patienta negative outlook with respect to possible complications. However, as can be seen, a single collective characteristic value may produce outliers, and the actual difference between the Tand Tvalues may be large in some cases and small in other cases, limiting their predictive value in individual cases. This can be why it may be advisable to perform the analysis for several different collective characteristics and feed them into the decision model.

6 FIG. 5 FIG. In, a table with outcomes of a perfusion based complication prediction can be shown. In this case, the outcomes for three different collective characteristics can be shown for the same data pool of patients, together with a mean ratio of the three characteristics and the final prediction for each patient. These collective characteristics comprise the maximum slope during the ingress phase that can be shown in, the maximum slope in the time frame from maximum intensity to 10 seconds after the maximum intensity, and the variance of the average slopes in the time frame from beginning of the ingress to 40 seconds after the beginning of the ingress.

8 9 3 1 2 8 9 5 FIG. For example, in the case of patientsand, the first value can be +7,736 and −3.901, respectively, which represent the fact that in, the Tpoint in each case can be a strong outlier with respect to the T-Tlines, one above, the other below. The values of +7,736 and −3,901 for patientsandcan be calculated as

1 2 which can be the formula that defines the Tvalue as +1 and the Tvalue as −1. The same formula can be used for the other values and patients as well.

In some cases, one of the three characteristic values could not be ascertained. In those instances, the mean ratio was calculated as the arithmetic middle of the remaining two values, in all other cases as the arithmetic middle of all three values. The final prediction of impaired or non-impaired aortic blood supply to the sigmoid colon depended on whether the mean ratio came out positive or negative.

All named characteristics, including those taken from the drawings alone, and individual characteristics, which are disclosed in combination with other characteristics, are considered alone and in combination as important to the disclosure. Embodiments according to the disclosure can be fulfilled through individual characteristics or a combination of several characteristics. Features which are combined with the wording “in particular” or “especially” are to be treated as preferred embodiments.

10 Clinical decision support system; 12 computer; 14 camera; 16 illumination device; 18 controller; 20 database; 22 user interface device; 30 1 normalized Tfluorescence signals; 32 2 normalized Tfluorescence signals; 34 3 normalized Tfluorescence signals; 36 negative control signals; 100 method; 102 1 2 3 load T, T, TICG perfusion image time series from database; 104 1 2 3 normalize T, T, TICG perfusion image time series; 106 define regions of interest; 108 1 2 3 extract T, T, TICG fluorescence intensity time series from the regions of interest; 110 1 2 3 extract collective characteristics from the T, T, TICG fluorescence intensity region of interest time series; 112 3 1 2 evaluate proximity of extracted Tcollective characteristics to extracted Tand Tcollective characteristics; 114 calculate and output a complication prediction; 200 method for determining collective characteristics; 202 1 2 load pre-normalized T, TICG fluorescence intensity time series for multiple patients from database; 204 extract variety of collective characteristics from ICG fluorescence intensity time series; 206 1 2 perform statistical analysis to determine those collective characteristics that are most significantly different for Tand Tfluorescence intensity time series, respectively; 208 output resulting collective characteristics definitions.